The present invention relates generally to natural language processing, and more specifically to identifying problematic language in inclusion and exclusion criteria.
Many business domains apply criteria to candidates, such as customers, patients, applicants, etc. to check their eligibility for a given service. For example, the criteria may define characteristics required for services such as inclusion in a medical trial, eligibility for insurance, enrolment in an activity, etc. These criteria may be written in terms of inclusion, i.e. the candidate must exhibit these characteristics, and exclusion, i.e. the candidate must not exhibit these characteristics. Often times, however, this criteria is written in a confusing or negating manner which can be difficult to decipher. As computing systems begin to process more natural language criteria sets, it is helpful to clarify criteria with problematic language.